The invention discloses a multi-test scene analysis and number reduction method for autonomous vehicle, which comprises the following steps: determining a group of scenes needing to be tested, parameterizing each scene, completing test track analysis under a specific scene, and obtaining a track of an uncompleted task; on the basis of the track, analyzing parameters related to the automatic driving capability, and defining the parameters as scene characteristic parameters; and after the analysis of a plurality of scenes is completed, obtaining a multi-scene characteristic parameter group, andremoving overlapped or contained parameter groups in the group, namely correspondingly removing the scene. The automatic driving vehicle multi-test scene analysis and number reduction method can be used for reducing the number of scenes needing to be tested in various existing automatic driving vehicle test and evaluation processes, so that the test quantity is reduced, the test cost is reduced, and the development of the automatic driving technology is promoted.